Slashdot Mirror
FDA Approves Vaccine For Prostate Cancer
reverseengineer writes "The US Food and Drug Administration has given its first first approval for a therapeutic cancer vaccine. In a clinical trial 'involving 512 men, those who got Provenge (sipuleucel-T) had a median survival of 25.8 months after treatment, while those who got a placebo lived a median of 21.7 months. After three years, 32 percent of those who got Provenge were alive, compared with 23 percent of those who got the placebo. ... "The big story here is that this is the first proof of principle and proof that immunotherapy works in general in cancer, which I think is a huge observation," said Dr. Philip Kantoff, chief of solid tumor oncology at the Dana-Farber Cancer Institute in Boston and the lead investigator in Dendreon's largest clinical trial for the drug. "I think this is a very big thing and will lead to a lot more enthusiasm for the approach."'"
According to my calculation, if the null hypothesis were true (i.e. the vaccine were just a placebo) there would be about a 1 in a million chance of a result this extreme (4.85 standard deviations above the mean). So it is highly significant statistically speaking. Whether it is clinically significant or not is a different question, of course.
Several major points:
1. There were 512 people in the trial. Assuming that these were split into two groups, that n=256. Of course it depends on the standard deviation, but you could get your p value very low with an n like that.
2. Human trials of drugs that treat potentially fatal conditions are generally only allowed for patients who've failed "best available" therapy at least once, because it would be unethical to deny standard therapy to someone in a trial. The relative risk reduction may get better when used outside the context of the trial because of that. Or not. Remains to be seen, as usual.
3. As our understanding of the immune system and the molecular processes underlying cancer improves, we will slowly unravel a huge potential for case-by-case-based treatment of cancer. As a matter of fact, I believe a recent study I am too lazy to look up to link has shown that people whose cancer therapies relied on analyzing the biochemistry of each individual tumor resulted in about a 50% increase in 5-year survival, compared to conventional pathology-based treatments. The future is bright for oncology.
4. The caveat of #3 is that such treatments are expensive, and will get progressively more expensive based on the degree of testing and individualization required (until the wide use and technology make them cheaper of course). This will necessarily introduce a further divide into the available treatments for the rich and the poor, and contribute to the class struggle that's already rather inflamed. The problem is that there's no OTHER way - giving everyone $100'000 treatments would bankrupt us rather quickly. Instead, similar to the case with electronics, we will simply have to suffer through the period of expensive first-adopter treatments, until the improvements in laboratory techniques and high-throughput testing make such treatments increasingly affordable.
If you assume each person has (independently) a 0.23 probability to survive 5 years, then the overall distribution must be binomial with standard deviation sqrt(n*p*(1-p)).